Multivariate Calibration of Near Infrared Spectroscopy in the Presence of Light Scattering Effect: A Comparative Study

Abstract

Abstract:When analyzing heterogeneous samples using spectroscopy, the light scattering effect introduces non-linearity into the measurements and deteriorates the prediction accuracy of conventional linear models. This paper compares the prediction performance of two categories of chemometric methods: pre-processing techniques to remove the non-linearity, and non-linear calibration techniques to directly model the non-linearity. A rigorous statistical procedure is adopted to ensure reliable comparison. The results suggest that optical path length estimation and correction (OPLEC) and Gaussian process (GP) regression are the most promising among the investigated methods. Furthermore, the combination of pre-processing and non-linear models is explored with limited success being achieved.

Chemometric determination of the length distribution of single walled carbon nanotubes through optical spectroscopy

“…Prior to developing the non-linear calibration models (ANN and GPR), PLS is first applied to reduce the dimension of the data, which is a common method used to reduce the computational cost of parameter estimation in non-linear models [34]. Conceptually, non-linear models are capable of direct modelling of the non-linearity; we also empirically observed that pre-processing methods, which aim to remove the non-linearity of the data, do not improve the non-linear calibration models [25]. As a result, these scattering correction techniques are not used for nonlinear models.…”

“…wavelength-dependent multiplicative effect due to disturbances). In the literature, non-linear calibration has shown comparable prediction performance with pre-processing techniques [25].…”

“…This is usually the result of the difference in sample homogeneity, the temperature variation during the analysis, and so on, giving rise to non-linear spectral data. In the literature, baseline and other noises can be addressed by using pre-processing methods and/or non-linear calibration models [25]. Typical pre-processing methods to remove non-linearity in the spectral data include first (D1) and second derivatives (D2) [26,27], standard normal variate (SNV) [28], extended multiplicative signal correction (EMSC) [29], and extended inverted signal correction (EISC) [30].…”

Near-Infrared Spectroscopy Combined with Multivariate Tools for Analysis of Trace Metals in Environmental Matrices

“…The ANN on the other hand is a flexible mathematical structure capable of identifying complex Developments in Near-Infrared Spectroscopynonlinear interactions between input and output data sets. This method is reported to be useful and efficient, especially in problems for which the characteristics are difficult to describe using physical equations [53]. The ANN model has been shown to perform better than linear models [54].…”

Meta-modelling for fast analysis of CFD-simulated vapour cloud dispersion processes

“…Second, the meta-model can have different forms, such as polynomial functions, neural networks, and Gaussian process regression (GPR, also known as kriging) (Palmer & Realff 2002, Chen et al 2006. Of particular focus in this work is GPR, for its reliable prediction results when compared with alternative methods (O'Hagan 2006, Wang et al 2011.…”